Me too. I asked how to auto set 0 phase = 0 dB a while back and no correct answer. The auto scale is rubbish. Not too hard to do manually however.
Yes, the Tian probe is symmetrical. Usually seen as a convenience that you can't stuff it up.
Would it work for you to add or subtract 180 to the result and call it "Phase Margin"?
The extra free parameters allow a better result for TMC in theory.
Whether that makes a difference in practice I want to learn.
OIC is shorter than "pure Cherry"! and when someone finds an earlier reference then we don't all have to rename it.
Edmond, in his modesty,
didn't call TMC after himself, so did not have to climb down when the earlier patent turned up.And descriptive names are easier to learn too.
Best wishes
David
modesty
As a matter of fact, It was Bob Cordell who inspired me to baptize my baby 'TMC'. He was the first one who used the word 'transition', describing the compensation changeover between OPS and TIS.
Cheers,
E.
[...]
Edmond, in his modesty,
And descriptive names are easier to learn too.
Best wishes
David
As a matter of fact, It was Bob Cordell who inspired me to baptize my baby 'TMC'. He was the first one who used the word 'transition', describing the compensation changeover between OPS and TIS.
Cheers,
E.
Well this is why I started this thread. Anyone care to use the extra free parameters to approach the 'pure Cherry' performance? Only 20dB THD to catch up on.
There's been a lot of pontificating about poles & zeros etc but its the final result that matters. How simple? How stable? What distortion? Will this hold in 'real life' with different loads & devices?
The one I want gives the best results for these and I dun care whether its got apple, oranges, poles or zeros inside.[*]
Dem be not descriptive names but TLAs which are even more obfuscating to the uninitiated unwashed masses en piple luk mi hu kunt reed en riteOIC is shorter than "pure Cherry"! and when someone finds an earlier reference then we don't all have to rename it.
Edmond, in his modesty,
And descriptive names are easier to learn too.
eg waz iz TIS? I gather its a VAS but wat it mean?
[deleted: 12 pg rant why TMC should be called Baxandall's compensation cos he was the first to explain the need and analyse it in detail ... even if he wasn't the first to use it bla bla ... and it isn't better than 'pure Cherry' .. but he's a guru ... etc .. and I've scored more pages of rant than you even though they've been deleted ... ]
[*]OK, I confess. I do want to know about poles, zeros, apples etc but only if they help me meet those targets. They mean nothing in themselves.
Last edited:
Well this is why I started this thread. Anyone care to use the extra free parameters to approach the 'pure Cherry' performance? Only 20dB THD to catch up on.
When I said "I want to learn" I didn't mean "I want someone to work it out for me"
I have started to work on this. You have only showed the phase for one of the loops in each case. As was learned earlier, this can be deceptive.
In OIC case it is the Output loop that is critical.
Even for the outermost loop the OIC phase is worse.
More soon.
Best wishes
David
TIS = Trans Impedance Stage = Current in, Volts out. It is a descriptive TLA
I have started to prefer just IS = Intermediate Section. Makes no presumption about how the amp will eventually work.
Minor layout recommendation. It would be nice if you turned of the grid for the ASC pictures. Less clutter on the screen.
Last edited:
This isn't the first time I've come across this gentlemen's agreement not to name poor amplifiers. What's the background? No doubt it won't look quite so **** weak once explained.
On the topic, Cherry had a few recommendations for successful/stable use of Pure Cherry/OIC which are listed in his 1997 Electronics World article 'Ironing Out Distortion'. I don't have that article with me now, but I remember that capacitors from the bases of the output drivers to the rails was one of the recommendations. I don't see these in the schematic of post 4. They were included in the calculation that showed the VAS emitter resistor aiding stability (1982 JAES article).
Ricardo, Pure Cherry won't catch on as a name, I'm afraid. Doesn't have enough letters in the initialism.
Edit: The irony... **** weak censored. But at least I've learned that I had the swearing hierarchy wrong: kunt allowed; **** weak not allowed.
Last edited:
A different Andrew's view,
the Zobel before the output inductor is there to provide an HF load for the amplifier. That HF load MUST have low inductance, else it can't effectively load the output stage @ HF.
That low inductance requires the Zobel to be closely coupled around the output stage devices and the HF decoupling attached to the output devices.
One CANNOT achieve HF loading capability, if the Zobel is remote from the output stage.
Last edited:
Yes, a Zobel after the inductor requires the inductor to be on the board and a non-inductive damper resistor.
But Cherry discussed both possibilities as do-able, with some benefits each way. .
I am inclined to a distributed Zobel before the inductor and a pi circuit, with some capacitance after as well.
Never had an answer from Mr Cordell about whether to connect the distributed Zobel directly to each output emitter (and thus before the emitter resistors rather than after as is usually done). Makes sense to me, you want to jump in on that question too?
Best wishes
David
Last edited:
I asked this question on Pi (distributed) in the other Thread, but the sole response seem to discard it as a rubbish question. No one else has picked up on the different aspects I enquired about.
We both seem to have arrived at this independently and it makes sense to Edmond too.
Three is a quorum*, I'm convinced.
* or the start of a joke (A Scotsman, a Dutchman and an Australian all...)
Best wishes
David
My opinion is that comparison between TPC and TMC with the same capacitances and resistor value will not show real differences. Why not to simulate a real amp and set the compensation value to get the same phase and gain margin and the similar result with square wave test on different capacitive load(yet better to simulate real loudspeaker and line parameters.
First equal C will not show the best of both compensation, also important is what side to put bigger cap(not the same for both compensation).
When I simulated TMC vs’ TPC with above requirement the result with TMC was a bit better then with TPC. http://www.diyaudio.com/forums/solid-state/188133-tmc-tpc-my-dilemma-resolved.html
BR Damir
First equal C will not show the best of both compensation, also important is what side to put bigger cap(not the same for both compensation).
When I simulated TMC vs’ TPC with above requirement the result with TMC was a bit better then with TPC. http://www.diyaudio.com/forums/solid-state/188133-tmc-tpc-my-dilemma-resolved.html
BR Damir
In your OIC there is about the same ULGF wrapped around your IS + OPS as the TPC/TMC has around just the IS.
And the OPS transistors are about one twentieth the speed?
That doesn't look "cherry" to me
Best wishes and off to bed for me
David
I've found (in simulation) one reason you might get oscillations on part of a waveform. Full story is here. Short version:
A lot of folks test their amp's stability by doing AC analyses (or other tests) under quiescent DC conditions. The problem is that most output devices show a large drop in Ft and beta at high currents. This can cause a significant change in stability margins, or even outright instability. A classic symptom is bursts of oscillation triggered by clipping, as shown below.
I think you're walking into the same trap. Here's a simple experiment: Try redoing the AC analyses of the 3 circuits with a 6A current source connected to the output. Or to be more realistic, just connect a 4 Ohm load to the output, and change the DC conditions to get +25V or +30V at the output.
I'd be interested to compare the results to what you showed in posts 1, 2 and 4.
but I don't see the shunt cap here http://www.diyaudio.com/forums/solid-state/235188-tpc-vs-tmc-vs-pure-cherry.html#post3474671
Maybe this should be reserved for cases where Cherry's recommendations have been followed purely. From Electronics World July 1997:
I can believe in local parasitic oscillation of the first member of the output-stage
Darlington — the ‘drivers‘ - as distinct from oscillation of the main feedback loop. Driver transistors such as BD139/140 with fT around 100MHz usually oscillate when biased to 5-10mA at the end of 10-20cm leads. Check the frequency of the oscillation: is it near the unity-loop-gain frequency. or is it significantly higher? For a common-collector-output amplifier with the low-level stages shown in Fig. 4.
[Formula for unity-loop-gain frequency doesn't copy well]
My amplifiers always feature impeccable layout and bypassing with separate quiet and noisy ground tracks, and short leads to the drivers. All of this discourages parasitics.
For the same reason I routinely provide ‘stopper’ capacitors of 30~50pF between collector and base of the drivers, using the shortest possible leads — no more than 1cm.
My amplifiers always incorporate a correctly-designed load-stabilising network.
My amplifiers always include judicious emitter degeneration in the second stage as
shown in Fig. 6.
In common-collector amplifiers — as distinct from my preferred common-emitter —I use a bootstrapped resistor as the second-stage current source.
I only show 'Loop Gain' cos several people are pontificating about how this is the answer to Life, the Universe & Everything.
Loop Gain, Bode, Nyquist, Tian etc are all 'linear measures' and are in fact the least useful methods to assess stability in da 21st century. But apart from their historical value to pedants, they are useful if you are trying to tweak an amp for better stability. But you certainly don't rely on them to for a Yah or Nay answer.
As an example, the true 'linear' test of stability is whether the Nyquist plot (which I've just found LTspice can do!) encircles (-1,0). All the Bode/Tian etc stuff only attempts a crude view of Nyquist.
It's actually very difficult to fudge a good amplifier so Nyquist encircles (-1,0). You have to do really nasty things to the compensation & other stuff. It's quite easy to get huge peaks on the Closed Loop response though and if you look at the Loop Gain in Nyquist, you see it approaches (-1,0) easily. But getting it to encircle (-1,0) is much more difficult.
In post #1, I explain how I assess stability, both virtual in SPICE world and also the 'real world'.
The major tool in LTspice is .TRANS with loadsa different loads and some different BJT models too.Bob's chapter on how to assess stability is excellent and the most important thing you should come away with is DON'T TRUST ANY SINGLE STABILITY CRITERIA.
But in fact, the easiest test of 'linear' stability is to plot Close Loop gain and check for peaking. It's less useful if you are trying to improve stability. LTspice Square Wave response uses .TRANS so isn't a 'linear' measure.
I did name one when its designer extolled its virtues on this forum. The 'gentleman' (?!?)
immediately disowned it and I was banned Please post a copy of Electronic World 1997 article if you find it.)
HERESY WARNING!
In Jurassic times, I amended Cherry's 1982 matrix analysis to take these out. (I cud reed en rite in dem daz.)
Unlike da gurus, I like the VAS collector to remain at High Z. The input of a Class B output stage has to swing VERY quickly to get across the evil crossover region. Anything that slows it down there is evil. That be why 'pure Cherry' is better than evil TMC, etc.
But not allowing Ccb on the drivers does remove one tool against 'parasitic' instability. The 'cure' includes tighter layouts (NO '10-20cm leads') and careful choice of the cap across the 'bias spreader'. Also useful is C9 across the VAS emitter resistor.
Plis dun mek fun orf mi. Us pipul hu kunt reed en rite r sensitive bout it.
Damir, will you try my .ASC files and suggests values for better performance from the TPC & TMC version? I've tried to make it as easy as possible for people to do this. I don't care what 'margins' etc are .. just what stability the final amp has.dadod said:
As I've said ad nauseum, my stability criteria isn't a 'linear' measure like Phase/Gain Margin/Nyquist .. but what happens on different loads on both Close Loop .ASC and .TRANS in SPICE world .. and the same in the 'real world'.
This is what the outside 'real world' sees.
I think my models are quite close (??) to 'real' amps .. at least for 'stability'. I need to take it one stage further and add some stuff for PSU.
I don't think you have that quite right but see above re: stability criteria.Dave Zan said:
The same transistors are used in all 3 versions.
One reason for my respect for LTspice is that .TRANS actually simulates this very well.godfrey said:
I don't think you'll see my 3 examples exhibit this cos I check for it. But if you modify them to demonstrate supa dupa THD, please check for it too.
BTW, it was da Golden Pinnae amps which did this .. not my designs, sim or 'real world'.
Yes. This helps but then loses one of the major virtues of 'pure Cherry'. They are not strictly necessary. If used, they need to be AT the physical drivers.
Prof. Cherry offers his VAS emitter resistor. IMLE, adding a judicious zero by decoupling this is very useful too.
You also have to consider what's within the 'pure Cherry' loop. In my post #4 example, we have an enhanced VAS (2 devices & roll-offs) and EF2 (another 2). You can ALWAYS stabilise 3 roll-offs if you reduce input transconductance (with R22) sufficiently though the resulting amp may not have very good THD. Four roll-offs is in "tweak it" territory.
Five, as favoured by Bob with his triple outputs, may be impractical if we want to retain simplicity. (?!!
)But astx shows it is possible to get excellent results with similar simplicity to post #4. I would personally like to see even better performance with even less complexity.
[deleted: zillion pg rant .. 'descriptive TLA' is an oxymoron and evil ... should be banned and their perpetrators de-pinnaed .... ]
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.